Typical Workflow to Integrate Oracle Analytics with Generative AI Models

Perform these tasks to integrate Oracle Analytics with generative AI models sourced from Oracle Cloud Infrastructure (see Option A below) or Oracle Autonomous AI Lakehouse (see Option B below).

Task	Option A Oracle Cloud Infrastructure Generative AI	Option B Oracle Autonomous AI Lakehouse
1. Select a model that suits your requirements.	To use On-Demand models that're already present in your OCI tenancy, make them accessible in the OCI compartment where you will use them (see Step 2 below). To use Dedicated models, you need a dedicated cluster. After provisioning, you can deploy an Oracle foundational model or your own custom model via a private endpoint. See Adding a Model to a Private Endpoint.	Your Generative AI Model and your datasets must reside in the same Autonomous AI Lakehouse database. Requires creating an AI profile in the database.
2. Set up your model.	Setup Oracle Analytics with OCI permissions to use OCI GenAI: : Create required OCI IAM policies to give OAC access to Generative AI. See Policies Required to Integrate Oracle Analytics with Generative AI Models. If you're connecting to OCI using Resource Principal, provision users with permissions to use OCI Generative AI. See Provide Users With Security Permissions to use Generative AI Models.	GRANT execute on DBMS_CLOUD_AI to <user>. See Perform Prerequisites for Select AI. Create your AI profile in the database. See DBMS_CLOUD_AI documentation (CREATE_PROFILE procedure). Important constraints: Datasets that you process and generative AI models must be from the same Oracle Autonomous AI Lakehouse. Datasets must be accessed via the same database connection.
3. Create a Oracle Analytics connection to your model source.	Create a Connection to Your Oracle Cloud Infrastructure Tenancy If your connection to OCI uses an API key, due to the latency between creating an API Key in OCI and it being active in a tenancy, wait several minutes before trying to create the connection. When you save your connection, choose a location in the Shared Folders area (required for execution of AI functions by end users). See Share Your Content.	Connect to Oracle Autonomous AI Lakehouse When you save your connection, choose a location in the Shared Folders area (required for execution of AI functions by end users). See Share Your Content.

Task

Option A Oracle Cloud Infrastructure Generative AI

Option B Oracle Autonomous AI Lakehouse

1. Select a model that suits your requirements.

To use On-Demand models that're already present in your OCI tenancy, make them accessible in the OCI compartment where you will use them (see Step 2 below).
To use Dedicated models, you need a dedicated cluster. After provisioning, you can deploy an Oracle foundational model or your own custom model via a private endpoint. See Adding a Model to a Private Endpoint.

Your Generative AI Model and your datasets must reside in the same Autonomous AI Lakehouse database.
Requires creating an AI profile in the database.

2. Set up your model.

Setup Oracle Analytics with OCI permissions to use OCI GenAI: :

Create required OCI IAM policies to give OAC access to Generative AI. See Policies Required to Integrate Oracle Analytics with Generative AI Models.
If you're connecting to OCI using Resource Principal, provision users with permissions to use OCI Generative AI. See Provide Users With Security Permissions to use Generative AI Models.

GRANT execute on DBMS_CLOUD_AI to <user>. See Perform Prerequisites for Select AI.
Create your AI profile in the database. See DBMS_CLOUD_AI documentation (CREATE_PROFILE procedure).

Important constraints:

Datasets that you process and generative AI models must be from the same Oracle Autonomous AI Lakehouse.
Datasets must be accessed via the same database connection.

3. Create a Oracle Analytics connection to your model source.

Create a Connection to Your Oracle Cloud Infrastructure Tenancy

If your connection to OCI uses an API key, due to the latency between creating an API Key in OCI and it being active in a tenancy, wait several minutes before trying to create the connection.

When you save your connection, choose a location in the Shared Folders area (required for execution of AI functions by end users). See Share Your Content.

Connect to Oracle Autonomous AI Lakehouse

When you save your connection, choose a location in the Shared Folders area (required for execution of AI functions by end users). See Share Your Content.

When you've completed steps 1 to 3 above, you can register and use Generative AI models by completing steps 4 to 8.

Task	Steps
4. Register a model.	Make a Generative AI Model Available in Oracle Analytics
5. Grant author permissions.	Share a Generative AI Model With Other Users in Oracle Analytics You should grant your authors' user/role Read-Only or higher access, so that they can create content with your Generative AI models. If you are the model owner, you have Full Control permission over the model. You can optionally delegate Full Control of your model to other users, so they can grant access to other users and roles.
6. Create content using generative AI functions.	Authors with author permissions can now create content with Generative AI functions in these areas: Data Flows - see Apply Generative AI In a Data Flow in Oracle Analytics. Workbooks - see Apply Generative AI in a Workbook in Oracle Analytics. Data Preparation - see Apply Generative AI in Data Preparation in Oracle Analytics.
7. Grant consumer permissions.	Share a Generative AI Model With Other Users in Oracle Analytics Permissions required: Users need at least Read-Only access in order to create content (for example, to use the Generative AI Models in workbooks). Users need at least Consume access in order to execute any queries using the AI Functions Users need Full Control to change or assign any of these permissions to other users or roles.
(Optional) 8. Migrate content.	If you migrate content to another Oracle Analytics instance: Generative AI Models are excluded from DVA Import/Export. Re-register models or make them available in the new instance to use the AI functions. Generative AI Models are included in BAR file exports alongside other objects in Oracle Analytics. Upon restoration using BAR, models can be accessed by any previous users if permissions were also exported. See Take Snapshots and Restore Information.